First, the main finding in our current study is the rejection of the pharmacokinetic mismatch hypothesis for emergence of isoniazid and rifampin resistance. Indeed, the isoniazid-resistant subpopulation was actually greater with perfect matching during the early time points, especially in the experiments with a drug-susceptible M. tuberculosis
inoculum. Rifampin monoresistance was rarely encountered. When systems were “rigged” with a subpopulation monoresistant to 5% rifampin and carrying an rpoB
codon 531 mutation, known to be stable and biofit (10
), this resistant subpopulation was nevertheless transient and was likely killed by isoniazid. Indeed, the proportions of drug-resistant isolates did not change during the studies for those HFSs that did not receive drug treatment, suggesting that biofitness was not a major factor. Even though the katG
mutant lasted longer in the HFS, this too was eventually eliminated. Thus, even with pre-existent resistance in a proportion of ≥1%, there was elimination of the drug-resistant subpopulations despite mismatch. Clearly, the drugs protect each other, an effect we document to be independent of mismatching.
Drug resistance evolves more commonly in patients with immunodeficiency. Thus, our HFS model, which lacks any immune system, should allow easy emergence of drug resistance. Indeed, during isoniazid and rifampin monotherapy, the HFS leads to emergence of drug resistance within 3 to 10 days (13
). As a result, the model has been criticized by other scientists for being too permissive in allowing the emergence of anti-TB-drug resistance compared to the scenario in patients (1
). Yet no drug resistance emerged with deliberate pharmacokinetic mismatch in the HFS, as long as both drugs were administered to each system. In addition, although two representative experiments are reported, four experiments were performed, none of which generated greater drug resistance with mismatch by the end of the experiments. Furthermore, we even used lower critical concentrations of rifampin and isoniazid than the standard ones to increase detection of drug resistance (7
). The most likely explanation is that pharmacokinetic mismatch does not lead to greater likelihood of M. tuberculosis
Pharmacokinetic mismatch has been well documented to lead to drug resistance in the treatment of other infectious diseases, which is especially important during antiretroviral therapy (4
). It has been estimated that in patients with AIDS, there is a mutant resistant to any one of the drugs at any one time, given the large number of virions and the legendary error-prone replication of HIV. In addition, HIV's doubling time is shorter by several orders of magnitude than the half-life of some antiretroviral agents. Thus, replication during effective monotherapy with such drugs as efavirenz, when nucleoside analogues are gone, easily leads to emergence of drug resistance (3
). On the other hand, the physiology of M. tuberculosis
and the pharmacokinetics of first-line anti-TB drugs make resistance emergence from pharmacokinetic mismatching less likely. M. tuberculosis
doubling times are at least 24 h, while rifampin and isoniazid have extremely short half-lives of 0.9 to 4 h (17
). Thus, these drugs are long gone by the time the bacteria replicates for a single round, let alone two rounds. Even use of rifapentine, which has a much longer half-life, is in the continuation phase of therapy against nonreplicating persistent bacilli with a doubling time that approaches infinity, so that rifapentine is long gone by the time the bacteria double.
The second important finding was that the more mismatched regimens were consistently associated with better and faster microbial killing than the perfectly matched regimens. This finding opens up the possibility that scheduling the administration of the two pivotal drugs using a particular sequence could be a new paradigm for accelerating microbial killing without increasing emergence of drug resistance. This concept is already used in treating cancers, in which drugs are sequenced according to when they act during the cell cycle. In the case of the two anti-TB drugs we studied, the reasons for better microbial killing are unclear, but there are several possibilities. First, the effect could be similar to what is seen in cancer chemotherapy, even taking into account the differences in mycobacterial cell division and mammalian cell cycle. In this regard, rifampin inhibits mRNA transcription, while isoniazid works during cell growth and division via inhibition of mycolic acid synthesis. Both drugs have half-lives that are much shorter than the division time of the bacteria, so that it is possible that they could work during different stages of the bacterial cell division. Second, our deliberate pharmacokinetic mismatching can be viewed as a “dose-scheduling” maneuver. Isoniazid and rifampin microbial killing are both linked to the ratio of the area under the concentration-time curve to MIC (AUC/MIC), while resistance suppression is linked to the ratio of peak concentration to MIC (Cmax
). In this scenario, as long as the same AUC/MIC is achieved, killing effect would be independent of dose schedule, while resistance suppression would be best with the 24-h-mismatched regimen in which Cmax
/MIC was double that of other regimens. Indeed, the 24-h mismatch was superior to all others for resistance suppression, particularly in the sterilizing-effect experiments, where the initial population had both isoniazid- and rifampin-resistant subpopulations in proportions above the clinically meaningful threshold of 1% (). However, this would not explain the superior microbial killing by the 6-h- and 12-h-mismatched regimens. Third, there could be pharmacodynamic antagonism between isoniazid and rifampin, which would be ameliorated if isoniazid was administered long after rifampin has started to work. Regardless of the explanation, however, our findings suggest that different dosing schedules and sequences of administration should be further studied for anti-TB drug combinations.
The importance of our findings is that alternative explanations must be sought for mechanisms of how drug resistance emerges during anti-TB therapy. This also means that solutions that rely on minimizing pharmacokinetic mismatch, such as fixed dose combinations, and the design of regimens that rely on better matching to close the monotherapy “window,” will likely be ineffective solutions for combating drug resistance.